Luminous remote sensing lunar image synthesis method
Technical Field
The invention belongs to the field of noctilucent remote sensing image processing, and particularly relates to a synthesis method of a noctilucent remote sensing lunar image.
Background
The luminous remote sensing data is as follows: the optical image of the earth surface is obtained at night by using a remote sensing satellite, and noctilucent data which is not influenced by sunlight and contains persistent light sources from cities, towns and the like is obtained. The noctilucent data not only has the advantages of periodicity, macroscopicity, objectivity and the like of remote sensing data, but also can objectively reflect the track of human activities and the development of regional economy through the extensive research of scholars at home and abroad. Therefore, the noctilucent remote sensing is widely applied to the aspects of urban development and evolution analysis, estimation of social and economic parameters such as GDP (graphics data processing), fire monitoring, fishery research and the like.
The current night light remote sensing data with global coverage comprises the National Defense Meteorological Satellite plan DMSP/OLS (Defense Metalogical Satellite Program/Operational linear system) and the American Polar orbit Satellite NPP/VIIRS (National Polar-orbital Partnership/visual infracted Imaging Radiometer Suite). The NPP/VIIRS inherits and optimizes the glimmer detection capability of the former and eliminates the problems of pixel saturation, overflow and the like, thereby being widely applied.
Because the noctilucent data has objectivity, all night lights on the ground can be obtained from the space, including residential house light sources, district street lamps, lighting lights on urban roads, lights of commercial districts and industrial districts, and the like. Moreover, DMSP/OLS and NPP/VIIRS data are released freely and can be downloaded directly through a network, so that a plurality of scholars at home and abroad dig out the application of the noctilucent remote sensing in different fields. From 1997, the scholars perform regression analysis on the total value of noctilucent remote sensing and national production and find that the noctilucent remote sensing and the national production have high correlation. The application field of the noctilucent remote sensing is continuously expanded, and the application field comprises the aspects of population density, power consumption, poverty and kini coefficient estimation, urbanization monitoring and evolution analysis, ecological environment and health effect research, fishery research and the like.
At present, the research is mostly to use annual noctilucence data published by the official to carry out macro-scale analysis, light influence from aurora, fire, ships and other temporary light sources is filtered through monthly data synthesis and screening, and global mosaic processing is carried out, so that socioeconomic factor estimation and economic activity monitoring can be carried out in a longer time sequence range. Whereas DMSP/OLS satellites have failed in 2013, the american earth observation commission has also announced in its official networks that production of NPP/VIIRS Night-light data monthly, annually synthetic products was stopped from 10 months in 2019, with Day/DNB (Day/Night Band) Night data provided only on a daily basis. At present, only the Payne research institute of the university of the Colorado mining of the United states provides NPP/VIIRS monthly synthetic products, but the updating and releasing of the products are delayed by about half a year, and the latest data analysis requirements cannot be met. Meanwhile, DNB data provided by the American Earth Observation Committee has certain image comparability although being subjected to on-satellite radiometric calibration during acquisition. However, because the noctilucent remote sensing data can be influenced by factors such as a satellite platform, a load sensor, the atmosphere and the moonlight, the original DNB radiation values of the same place acquired on different dates still have large fluctuation. Therefore, the application analysis research cannot be directly carried out by using the DNB data of the day, so that the using capability of the NPP/VIIRS data under the conditions of serious natural disasters, new crown epidemic situations and local wars is limited.
Disclosure of Invention
The invention provides a method for synthesizing a luminous remote sensing lunar image, aiming at the problems that the prior art can not provide moon in time, the annual luminous data and day and night DNB data have large fluctuation and can not be directly used under the conditions of serious natural disasters, new crown epidemic situations and local wars.
The method comprises the following steps:
the method comprises the following steps of firstly, acquiring NPP/VIIRS day noctilucent data of a certain month aiming at an area to be monitored; searching administrative division data corresponding to the region;
preprocessing day night light data, and cutting according to administrative division data to obtain night light remote sensing grid data of an area to be monitored;
the method comprises the following specific steps:
step 201, converting the geographical coordinate projection of the noctilucent data into WGS84 geographical coordinates, and resampling the spatial resolution of the data to 0.00416667 degrees;
step 202, inlaying noctilucent data according to administrative division vector data to ensure that an area to be monitored can be covered;
and 203, cutting the noctilucent data by taking the administrative division vector plane data as a mask, and extracting the noctilucent remote sensing grid data of the region to be monitored.
Step three, converting original DNB radiation values of all pixels in the noctilucent remote sensing raster data into pixel brightness values DN corresponding to the original DNB radiation values respectively to form a noctilucent remote sensing DN value raster image;
the conversion formula of the ith pixel is as follows;
DNi=Radi*109
wherein DNiThe value of the luminance of the picture element, Rad, representing the ith pixeliAnd representing the original DNB radiation value of the ith pixel in the original noctilucent remote sensing raster data. N, N represents the number of pixels in the noctilucent remote sensing grid data of the region to be monitored.
Opening a noctilucent remote sensing DN value grid image, and screening the noctilucent data corresponding to the K dates without clouds and with better quality through cloud judgment and visual interpretation;
the number of the noctilucent data is K, and K represents the number of days meeting the screening condition in the month.
Cloud judgment means: obtaining a cloud mask file by detecting a cloud area, and detecting noctilucent remote sensing DN value raster data by using a cloud mask tool; when the area of the cloud area exceeds 3%, the cloud area is deleted as invalid data and does not participate in subsequent calculation;
visual interpretation means: and (4) the noctilucent remote sensing DN value raster image subjected to cloud judgment is checked by human eyes, and the missing data is deleted as invalid data.
Selecting respective target reference areas from the screened noctilucent data of each date, and respectively calculating the relative radiation correction coefficient of the noctilucent data of each date according to the pixel brightness values of the reference areas;
the method comprises the following specific steps:
step 501, aiming at the noctilucent data of each date, selecting each noctilucent data center urban area according to administrative division vector data, and taking the area with the maximum image brightness value as a reference area;
step 502, calculating the Mean value C _ Mean of the pixel brightness values corresponding to each reference region respectivelyk;
k=1,2,...K;
Step 503, sorting the K brightness Mean values, and selecting a median C _ MeanmedAs a reference value;
step 504, utilize the Mean value C _ Mean of the brightness values of each pixelkAnd the median value C _ MeanmedCalculate C _ Mean for each datekA relative correction factor;
the relative correction factor calculation formula for the kth date is as follows:
ak=C_Meanmed/C_Meank
sixthly, carrying out sectional compensation processing on noctilucent data in the areas to be monitored on different dates by using the relative correction coefficients;
the method comprises the following specific steps:
step 601, opening the screened noctilucent data of K dates, and respectively calculating the Mean value Mean of all pixels in the monitoring areak;
Step 602, aiming at the noctilucent data of each date, the DN value and the Mean value of each pixel are calculatedkComparing, and compensating the noctilucent data of the date by using the relative correction coefficient;
the compensation means that: updating the DN value of each pixel in the noctilucent data; the calculation formula is as follows:
wherein DNki' indicating updated DN value of ith pixel in luminous data of kth date, DNkiRepresenting the original DN value of the ith pixel in the luminous data of the kth date.
Seventhly, carrying out monthly synthesis processing on the compensated noctilucent data of K dates by adopting an averaging method to obtain a monthly noctilucent remote sensing image;
DN of value of each pixel in moonlight remote sensing imagei' calculation is as follows:
the invention has the advantages that:
1) the moonlight synthetic night-sensing remote sensing data with 99.98% similarity to official release reference moonlight data images is obtained through a radiation correction method of sectional compensation processing based on the original daily NPP/VIIRS DNB radiation value data. From the perspective of objective evaluation of noctilucence application, the precision error of the provincial administrative region range average light intensity index is better than 0.92%, and the requirement of macroscopic economic analysis is met.
2) The method solves the problem of poor timeliness of updating NPP/VIIRS (non-uniform-power-point/quasi-near-zero-spectral-intensity) luminous lunar synthetic data, basically realizes the lunar synthetic luminous remote sensing data, and effectively improves the application capability of the luminous data in major emergencies such as new crown epidemic situations, natural disasters and the like.
Drawings
FIG. 1 is a flow chart of a method for synthesizing a luminous remote sensing lunar image according to the present invention.
FIG. 2 is a comparison between the lunar night luminous map synthesized in Jianghe and Zhejiang Shanghai region and official release data in this embodiment
Detailed Description
The present invention will be described in further detail and with reference to the accompanying drawings so that those skilled in the art can understand and practice the invention.
A luminous remote sensing lunar image synthesis method is shown in figure 1 and comprises the following steps:
the method comprises the following steps of firstly, acquiring NPP/VIIRS day noctilucent data of a certain month aiming at an area to be monitored; searching administrative division data corresponding to the region;
preprocessing day night light data, and cutting according to administrative division data to obtain night light remote sensing grid data of an area to be monitored;
the method comprises the following specific steps:
step 201, utilizing ENVI software to convert a coordinate system of noctilucent data into a WGS-84 geographical coordinate system, and resampling the spatial resolution of the data to 0.00416667 degrees;
step 202, inlaying noctilucent data according to administrative division vector data to ensure that an area to be monitored can be covered;
and 203, cutting the noctilucent data according to the administrative division vector data to obtain noctilucent remote sensing grid data of the area to be monitored.
Step three, converting the original DNB radiation value of each pixel in the noctilucent remote sensing raster data into corresponding pixel brightness value DN (digital number) respectively to form a noctilucent remote sensing DN value raster image;
the conversion formula of the ith pixel is as follows;
DNi=Radi*109
wherein DNiThe value of the luminance of the picture element, Rad, representing the ith pixeliAnd representing the original DNB radiation value of the ith pixel in the original noctilucent remote sensing raster data. N, N represents the number of pixels in the noctilucent remote sensing grid data of the region to be monitored.
Screening and processing the noctilucent data, and obtaining the noctilucent data corresponding to the K dates without clouds and with better quality through cloud judgment and visual interpretation;
the number of the noctilucent data is K, and K represents the number of days meeting the screening condition in the month.
Firstly, opening a noctilucent remote sensing DN value grid image, and screening out data covered by cloud through cloud judgment;
cloud judgment means: obtaining a cloud mask file by detecting a cloud area, and detecting noctilucent remote sensing DN value raster data by using a cloud mask tool; when the area of the cloud area exceeds 3%, the cloud area is deleted as invalid data and does not participate in subsequent calculation;
and then screening out data containing cloud coverage and missing through visual interpretation to obtain noctilucent data of K dates with better quality in the month.
Selecting respective target reference areas from the screened noctilucent data of each date, and respectively calculating the relative radiation correction coefficient of the noctilucent data of each date according to the pixel brightness values of the reference areas;
the method comprises the following specific steps:
step 501, opening the screened noctilucent data of K dates, selecting a central urban area of the noctilucent data of each date according to administrative division vector data, and taking an area with the maximum image brightness value as a reference area;
the size of the selected area is usually 5 x 5 pixels;
step 502, respectively calculating the Mean value C _ Mean of the pixel brightness values in each reference areak;
k=1,2,...K;
Step 503, sorting the K brightness Mean values, and selecting a median C _ MeanmedAs a reference value;
step 504, utilize the Mean value C _ Mean of the brightness values of each pixelkAnd the median value C _ MeanmedCalculating a relative correction coefficient of each date;
the relative correction factor calculation formula for the kth date is as follows:
ak=C_Meanmed/C_Meank
sixthly, carrying out sectional compensation processing on noctilucent data in the areas to be monitored on different dates by using the relative correction coefficients;
the method comprises the following specific steps:
step 601, opening the screened noctilucent data of K dates, and respectively calculating the Mean value Mean of all pixels in the monitoring areak;
Step 602, aiming at the noctilucent data of each date, the DN value and the Mean value of each pixel are calculatedkComparing, and compensating the noctilucent data of the date by using the relative correction coefficient;
the compensation means that: updating the DN value of each pixel in the noctilucent data; the calculation formula is as follows:
wherein DNki' indicating updated DN value of ith pixel in luminous data of kth date, DNkiRepresenting the original DN value of the ith pixel in the luminous data of the kth date.
Seventhly, carrying out monthly synthesis processing on the compensated noctilucent data of K dates by adopting an averaging method to obtain a monthly noctilucent remote sensing image;
DN of each pixel value of lunar night light remote sensing imagei' calculation is as follows:
examples
As shown in FIG. 2, the experiment of synthesizing the day-scale NPP/VIIRS luminous data in Shanghai province is described.
Firstly, downloading all daily NPP/VIIRS noctilucent data in 2019 and 1 month from the official network of the national ocean and atmosphere administration, wherein the data format is h 5; monthly synthetic data for month 1 in 2019 was also downloaded from the Payne institute.
And then, carrying out format conversion, preprocessing and cutting on the daily data according to the coordinate system and the resolution of the monthly data to finish the conversion from the original DNB radiation value to the DN value.
Qualified data are screened through cloud judgment and manual visual interpretation, and data with good quality in 5 th stage, such as 20200119, 20200120, 20200129, 20200130 and 20200131, are reserved. According to the administrative division vector data and the image brightness situation, the brightest area in the Shanghai city center is selected as a reference area, and the compensation coefficient of each date data is calculated by using the DN value mean value and the median value of the reference areas of 5 dates.
Then, the noctilucence data of Jianghu and Zhehu on 5 dates are relatively corrected.
Taking the data of 1 month and 20 days in 2020 as an example, the updated luminous data of 1 month and 20 days in 1 month is obtained by the method of segmented compensation processing through the conversion coefficient 1.067985 and the whole graph mean value 3.568964.
And finally, carrying out mean value processing on the updated noctilucent data of 5 dates, and synthesizing to obtain a monthly synthetic product of 2019 in 1 month.
The average night light intensity of relevant research surfaces can provide good regional economic condition estimation, so that 14 city-level samples are selected according to the principles of randomness, uniformity and different characteristics, three provincial samples of Zhejiang, Jiangsu and Shanghai are selected, and the average light intensity index is used for evaluating the synthesis precision, and the result is shown in table 1.
TABLE 1 accuracy of synthesized monthly luminous remote sensing data
The MEAN column represents the synthetic result of the method, the MEAN _1 column represents the synthetic data result of the Payne research institute, the error rate column represents the error rate, and the ABS rate column represents the absolute error rate. As can be seen from Table 1: in the synthesis result of the city-level region range, the absolute error is maximum in Taizhou city, the maximum error is 7.24%, the minimum error is in the Loqing city, and the minimum error is 0.31%; the mean absolute error was 3.21%.
In the monthly synthesis result of the provincial region range, the error of the average light intensity index is 0.92%, and the provincial region range is commonly used for carrying out application and analysis on social and economic parameters such as national production total value and the like on noctilucent data, so that the synthesis result of the method has higher practicability.
Meanwhile, the synthetic data and the data issued by the American institute of research are evaluated from the similarity index of the noctilucent remote sensing image, and the cosine similarity calculation of the two images is calculated by a gray level histogram method, and the result shows that the similarity reaches 99.98 percent and is consistent with the visual effect. Through evaluation, from two aspects of objective indexes and subjective evaluation, the monthly synthesis result obtained by the method provided by the invention has better consistency with the monthly result issued by an official part, and can meet the requirement of macroscopic economic analysis.